Summary

Chapter 33 Summary: Beyond 5G / 6G Vision

Key Points

• 1.

  6G Spectrum: FR3 to Sub-THz. The 6G spectrum strategy spans three tiers: FR3 upper mid-band (7--24 GHz) offering the best balance of coverage and capacity for wide-area deployments; W-band (75--110 GHz) and D-band (130--175 GHz) for ultra-wideband short-range links; and sub-THz (200--300 GHz) for extreme bandwidth applications. Atmospheric absorption peaks (60 GHz O$_2$, 183 GHz H$_2$O) partition the spectrum into transmission windows. WRC-23 and WRC-27 decisions will shape the regulatory framework, while coexistence with incumbent services (satellites, radar) requires dynamic sharing mechanisms.

• 2.

  Network Architecture: Open RAN, AI-Native, NTN. The 6G network architecture rests on three pillars. Open RAN (O-RAN) disaggregates base station functions into interoperable O-RU/O-DU/O-CU components with a RAN Intelligent Controller (RIC) enabling third-party optimisation via xApps. AI-native design embeds machine learning into every protocol layer — from neural receivers and learned CSI feedback to foundation models for wireless. Non-terrestrial networks integrate LEO satellites (1--4 ms delay), HAPS (quasi-stationary at 20 km), and UAVs for ubiquitous global coverage, with 3GPP Release 17+ providing NTN-NR support through timing advance and Doppler pre-correction.

• 3.

  Near-Field Communications and XL-MIMO. As arrays grow to hundreds or thousands of elements at 6G frequencies, the Rayleigh distance $d_R = 2D^2/\lambda$ can reach hundreds of metres, placing most users in the near field. Beamfocusing replaces beamsteering: energy is concentrated at a specific 3D point $(r, \theta, \phi)$ rather than just a direction, enabling range-domain user separation. Spatial non-stationarity — where different array elements see different scattering clusters — requires new channel models with per-element visibility regions and sub-array-based processing.

• 4.

  In-Band Full Duplex. Full-duplex transmission on the same time-frequency resource promises up to $2\times$ spectral efficiency by eliminating the half-duplex UL/DL time or frequency split. The fundamental challenge is self-interference (SI), which can exceed the desired signal by 100--130 dB. A three-stage cancellation chain — propagation domain (20--40 dB), analog domain (30--60 dB), digital domain (30--50 dB) — achieves 100--130 dB total suppression. Massive MIMO spatial SI suppression further relaxes the requirements. Key applications include IAB nodes, monostatic ISAC, and latency reduction.

• 5.

  Over-the-Air Computation (AirComp). AirComp exploits the superposition property of the wireless channel to compute aggregate functions (e.g., arithmetic mean for federated learning) in a single channel use, regardless of the number of devices $K$. Each device pre-equalises its transmission via channel inversion $\eta_k = \sqrt{p_0}/h_k$, so all signals arrive coherently aligned at the server. The MSE decreases as $1/(K^2 p_0)$, and the communication latency is $\mathcal{O}(1)$ vs $\mathcal{O}(K)$ for orthogonal access. The weakest-link power bottleneck is mitigated by truncated inversion, MIMO receivers, and RIS-assisted channel shaping.

• 6.

  Digital Twins, Ray Tracing, and ISAC. Site-specific wireless digital twins combine 3D geometry, ray-tracing propagation engines, and ML-based calibration to create high-fidelity virtual replicas of physical wireless environments. Applications include predictive beam management, proactive handover, network planning, and synthetic training data for AI. Integrated sensing and communication (ISAC) uses the same OFDM waveform for data transmission and radar-like sensing, with range resolution $\Delta R = c/(2B)$ and velocity resolution determined by the sensing frame duration. The digital twin serves as the unifying framework: sensing updates the twin, and the twin drives AI-based network optimisation.